The effect of model selection on cost-effectiveness research

A comparison of kidney function-based microsimulation and disease grade-based microsimulation in chronic kidney disease modeling

Shusuke Hiragi, Hiroshi Tamura, Rei Goto, Tomohiro Kuroda

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Background: Cost effectiveness research is emerging in the chronic kidney disease (CKD) research field. Especially, an individual-level state transition model (microsimulation) is widely used for these researches. Some researchers set CKD grades as discrete health states, and the transition probabilities between these states were dependent on the CKD grades (disease grade-based microsimulation, MSM-dg), while others set estimated glomerular filtration rate value which determines the severity of CKD as a main variable describing patients' continuous status (kidney function-based microsimulation, MSM-kf). MSM-kf seems to reflect the real world more precisely but is more difficult to implement. We compared the calculation results of these two microsimulation models to evaluate the effect of model selection on CKD cost-effectiveness analysis. Methods: We implemented simplified MSM-dg and MSM-kf emulating natural course of CKD in general, and compared models using parameters derived from an IgA nephropathy cohort. After checking these models' overall behavior, life-years, utilities, and thresholds regarding intervention costs below which the intervention is thought as dominant (V0) or cost-effective (V1) were calculated. In addition, one-way and probabilistic sensitivity analyses were performed. Results: With base-case parameters, the calculated life-years was shorter in MSM-dg (73.89 vs. 75.80 years) while the thresholds were almost equal (86.87 vs. 90.43 (V0), 132.29 vs. 146.25 [V1 in 1000 USD]) compared to MSM-kf. Sensitivity analyses showed a tendency of the MSM-dg to show shorter results in life-years. V0 and V1 were distributed by approximately ±100,000 USD (V0) and ± 150,000 USD (V1) between models. Conclusions: Estimated cost-effectiveness thresholds by both models were not the same and its difference distributed too wide to be ignored. This result indicated that model selection in CKD cost-effectiveness research has large effect on their conclusions.

Original languageEnglish
Article number94
JournalBMC Medical Informatics and Decision Making
Volume18
Issue number1
DOIs
Publication statusPublished - 2018 Nov 9

Fingerprint

Chronic Renal Insufficiency
Cost-Benefit Analysis
Kidney
Research
Cost of Illness
Health Transition
Costs and Cost Analysis
Glomerular Filtration Rate
Immunoglobulin A
Research Personnel

Keywords

  • Chronic kidney disease
  • Cost effectiveness analysis
  • Disease modeling
  • Health economics

ASJC Scopus subject areas

  • Health Policy
  • Health Informatics

Cite this

@article{ac22484c1ffd48bb9da18ff5856db4e6,
title = "The effect of model selection on cost-effectiveness research: A comparison of kidney function-based microsimulation and disease grade-based microsimulation in chronic kidney disease modeling",
abstract = "Background: Cost effectiveness research is emerging in the chronic kidney disease (CKD) research field. Especially, an individual-level state transition model (microsimulation) is widely used for these researches. Some researchers set CKD grades as discrete health states, and the transition probabilities between these states were dependent on the CKD grades (disease grade-based microsimulation, MSM-dg), while others set estimated glomerular filtration rate value which determines the severity of CKD as a main variable describing patients' continuous status (kidney function-based microsimulation, MSM-kf). MSM-kf seems to reflect the real world more precisely but is more difficult to implement. We compared the calculation results of these two microsimulation models to evaluate the effect of model selection on CKD cost-effectiveness analysis. Methods: We implemented simplified MSM-dg and MSM-kf emulating natural course of CKD in general, and compared models using parameters derived from an IgA nephropathy cohort. After checking these models' overall behavior, life-years, utilities, and thresholds regarding intervention costs below which the intervention is thought as dominant (V0) or cost-effective (V1) were calculated. In addition, one-way and probabilistic sensitivity analyses were performed. Results: With base-case parameters, the calculated life-years was shorter in MSM-dg (73.89 vs. 75.80 years) while the thresholds were almost equal (86.87 vs. 90.43 (V0), 132.29 vs. 146.25 [V1 in 1000 USD]) compared to MSM-kf. Sensitivity analyses showed a tendency of the MSM-dg to show shorter results in life-years. V0 and V1 were distributed by approximately ±100,000 USD (V0) and ± 150,000 USD (V1) between models. Conclusions: Estimated cost-effectiveness thresholds by both models were not the same and its difference distributed too wide to be ignored. This result indicated that model selection in CKD cost-effectiveness research has large effect on their conclusions.",
keywords = "Chronic kidney disease, Cost effectiveness analysis, Disease modeling, Health economics",
author = "Shusuke Hiragi and Hiroshi Tamura and Rei Goto and Tomohiro Kuroda",
year = "2018",
month = "11",
day = "9",
doi = "10.1186/s12911-018-0678-7",
language = "English",
volume = "18",
journal = "BMC Medical Informatics and Decision Making",
issn = "1472-6947",
publisher = "BioMed Central",
number = "1",

}

TY - JOUR

T1 - The effect of model selection on cost-effectiveness research

T2 - A comparison of kidney function-based microsimulation and disease grade-based microsimulation in chronic kidney disease modeling

AU - Hiragi, Shusuke

AU - Tamura, Hiroshi

AU - Goto, Rei

AU - Kuroda, Tomohiro

PY - 2018/11/9

Y1 - 2018/11/9

N2 - Background: Cost effectiveness research is emerging in the chronic kidney disease (CKD) research field. Especially, an individual-level state transition model (microsimulation) is widely used for these researches. Some researchers set CKD grades as discrete health states, and the transition probabilities between these states were dependent on the CKD grades (disease grade-based microsimulation, MSM-dg), while others set estimated glomerular filtration rate value which determines the severity of CKD as a main variable describing patients' continuous status (kidney function-based microsimulation, MSM-kf). MSM-kf seems to reflect the real world more precisely but is more difficult to implement. We compared the calculation results of these two microsimulation models to evaluate the effect of model selection on CKD cost-effectiveness analysis. Methods: We implemented simplified MSM-dg and MSM-kf emulating natural course of CKD in general, and compared models using parameters derived from an IgA nephropathy cohort. After checking these models' overall behavior, life-years, utilities, and thresholds regarding intervention costs below which the intervention is thought as dominant (V0) or cost-effective (V1) were calculated. In addition, one-way and probabilistic sensitivity analyses were performed. Results: With base-case parameters, the calculated life-years was shorter in MSM-dg (73.89 vs. 75.80 years) while the thresholds were almost equal (86.87 vs. 90.43 (V0), 132.29 vs. 146.25 [V1 in 1000 USD]) compared to MSM-kf. Sensitivity analyses showed a tendency of the MSM-dg to show shorter results in life-years. V0 and V1 were distributed by approximately ±100,000 USD (V0) and ± 150,000 USD (V1) between models. Conclusions: Estimated cost-effectiveness thresholds by both models were not the same and its difference distributed too wide to be ignored. This result indicated that model selection in CKD cost-effectiveness research has large effect on their conclusions.

AB - Background: Cost effectiveness research is emerging in the chronic kidney disease (CKD) research field. Especially, an individual-level state transition model (microsimulation) is widely used for these researches. Some researchers set CKD grades as discrete health states, and the transition probabilities between these states were dependent on the CKD grades (disease grade-based microsimulation, MSM-dg), while others set estimated glomerular filtration rate value which determines the severity of CKD as a main variable describing patients' continuous status (kidney function-based microsimulation, MSM-kf). MSM-kf seems to reflect the real world more precisely but is more difficult to implement. We compared the calculation results of these two microsimulation models to evaluate the effect of model selection on CKD cost-effectiveness analysis. Methods: We implemented simplified MSM-dg and MSM-kf emulating natural course of CKD in general, and compared models using parameters derived from an IgA nephropathy cohort. After checking these models' overall behavior, life-years, utilities, and thresholds regarding intervention costs below which the intervention is thought as dominant (V0) or cost-effective (V1) were calculated. In addition, one-way and probabilistic sensitivity analyses were performed. Results: With base-case parameters, the calculated life-years was shorter in MSM-dg (73.89 vs. 75.80 years) while the thresholds were almost equal (86.87 vs. 90.43 (V0), 132.29 vs. 146.25 [V1 in 1000 USD]) compared to MSM-kf. Sensitivity analyses showed a tendency of the MSM-dg to show shorter results in life-years. V0 and V1 were distributed by approximately ±100,000 USD (V0) and ± 150,000 USD (V1) between models. Conclusions: Estimated cost-effectiveness thresholds by both models were not the same and its difference distributed too wide to be ignored. This result indicated that model selection in CKD cost-effectiveness research has large effect on their conclusions.

KW - Chronic kidney disease

KW - Cost effectiveness analysis

KW - Disease modeling

KW - Health economics

UR - http://www.scopus.com/inward/record.url?scp=85056405188&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85056405188&partnerID=8YFLogxK

U2 - 10.1186/s12911-018-0678-7

DO - 10.1186/s12911-018-0678-7

M3 - Article

VL - 18

JO - BMC Medical Informatics and Decision Making

JF - BMC Medical Informatics and Decision Making

SN - 1472-6947

IS - 1

M1 - 94

ER -